Update app.py

app.py

@@ -1,83 +1,96 @@
+# # Set your Groq API key here or use environment variable
+# GROQ_API_TOKEN = os.getenv("groq_api")
+# client = Groq(api_key=GROQ_API_TOKEN)
+
 import os
+import ffmpeg
+import whisper
 import streamlit as st
-from sentence_transformers import SentenceTransformer, util
 from groq import Groq
-from PyPDF2 import PdfReader
-
-# Set your Groq API key here or use environment variable
-GROQ_API_TOKEN = os.getenv("groq_api")
-client = Groq(api_key=GROQ_API_TOKEN)
-
-# Initialize the SentenceTransformer model for embeddings
-retriever = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
-
-# Knowledge base (documents) and embeddings
-documents = []
-document_embeddings = None
-
-# Function to retrieve top relevant document
-def retrieve(query, top_k=3):  # Retrieve top 3 relevant documents
-    if document_embeddings is None:
-        return None
-    query_embedding = retriever.encode(query, convert_to_tensor=True)
-    hits = util.semantic_search(query_embedding, document_embeddings, top_k=top_k)
-    top_docs = [documents[hit['corpus_id']] for hit in hits[0]]
-    return ' '.join(top_docs) if hits[0] else None  # Concatenate the top documents
-
-# Function to generate response using Groq
-def generate_response(query, context):
-    # Limit context size to prevent exceeding token limits
-    max_context_length = 200  # Adjust this number based on your needs
-    if len(context.split()) > max_context_length:
-        context = ' '.join(context.split()[:max_context_length])  # Truncate context to the first N words
 
+# Set the title and description of the app
+st.title("Audio/Video Transcription and Summarization")
+st.write("Upload your audio or video file, and this app will transcribe the audio and provide a summary of the transcription.")
+
+# Retrieve the API key from environment variables or Streamlit secrets
+GROQ_API_KEY = os.getenv("GROQ_API_KEY") or st.secrets["GROQ_API_KEY"]
+os.environ["GROQ_API_KEY"] = GROQ_API_KEY
+
+# Create a temporary directory if it does not exist
+temp_dir = "temp"
+os.makedirs(temp_dir, exist_ok=True)
+
+# Upload the audio or video file
+uploaded_file = st.file_uploader("Choose an audio or video file...", type=["mp4", "mov", "avi", "mkv", "wav", "mp3"])
+
+# Function to extract audio from video
+def extract_audio(video_path, audio_path="temp/temp_audio.wav"):
+    """Extracts audio from video."""
+    try:
+        # Run ffmpeg command with stderr capture for better error handling
+        ffmpeg.input(video_path).output(audio_path).run(overwrite_output=True, capture_stdout=True, capture_stderr=True)
+    except ffmpeg.Error as e:
+        st.error("FFmpeg error encountered: " + e.stderr.decode())
+    return audio_path
+
+# Function to transcribe audio to text using Whisper model
+def transcribe_audio(audio_path):
+    """Transcribes audio to text using Whisper model."""
+    model = whisper.load_model("base")  # Load the Whisper model
+    result = model.transcribe(audio_path)
+    return result["text"]
+
+# Function to summarize text using Groq API
+def summarize_text(text):
+    """Summarizes text using Groq API."""
+    client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
     response = client.chat.completions.create(
-        messages=[{
-            "role": "user",
-            "content": f"Context: {context}\nQuestion: {query}\nAnswer:"
-        }],
-        model="gemma2-9b-it"
+        messages=[{"role": "user", "content": f"Summarize the following text: {text}"}],
+        model="llama3-8b-8192"
     )
-    return response.choices[0].message.content.strip()
-
-# Function to handle PDF upload and text extraction
-def extract_text_from_pdf(file):
-    pdf_reader = PdfReader(file)
-    text = ""
-    for page in pdf_reader.pages:
-        if page.extract_text():
-            text += page.extract_text() + "\n"
-    return text.strip()
-
-# Function to update knowledge base with new content from PDF
-def update_knowledge_base(pdf_text):
-    global documents, document_embeddings
-    documents.append(pdf_text)
-    document_embeddings = retriever.encode(documents, convert_to_tensor=True)
-
-# Streamlit app layout
-st.title("RAG-based PDF Question Answering App")
-st.write("Upload a PDF, ask questions based on its content, and get answers!")
-
-# Upload PDF file
-uploaded_file = st.file_uploader("Upload a PDF file", type="pdf")
-if uploaded_file:
-    pdf_text = extract_text_from_pdf(uploaded_file)
-    if pdf_text:
-        update_knowledge_base(pdf_text)
-        st.success("PDF content successfully added to the knowledge base.")
-    else:
-        st.warning("No text could be extracted from the PDF.")
-
-# Question input
-question = st.text_input("Enter your question:")
-if question:
-    retrieved_context = retrieve(question)
-    if retrieved_context:
-        answer = generate_response(question, retrieved_context)
-        st.write("Answer:", answer)
+    summary = response.choices[0].message.content
+    return summary
+
+# Complete function to process audio or video
+def process_media(media_file):
+    """Processes audio or video: extracts audio, transcribes it, and summarizes the transcription."""
+    # Save the uploaded file to a temporary path
+    temp_file_path = os.path.join(temp_dir, media_file.name)
+    with open(temp_file_path, "wb") as f:
+        f.write(media_file.getbuffer())
+
+    # Determine if the file is a video or audio based on the file extension
+    if media_file.name.endswith(('.mp4', '.mov', '.avi', '.mkv')):
+        # Step 1: Extract audio from video
+        audio_path = extract_audio(temp_file_path)
     else:
-        st.write("I have no knowledge about this topic.")
+        audio_path = temp_file_path  # If it's already audio, use it as is
+
+    # Step 2: Transcribe audio to text
+    transcription = transcribe_audio(audio_path)
+    st.write("### Transcription:")
+    st.write(transcription)
+    
+    # Step 3: Summarize transcription
+    summary = summarize_text(transcription)
+    st.write("### Summary:")
+    st.write(summary)
+    
+    # Clean up temporary files if needed
+    os.remove(temp_file_path)
+    if media_file.name.endswith(('.mp4', '.mov', '.avi', '.mkv')):
+        os.remove(audio_path)
+
+# Run the app
+if uploaded_file is not None:
+    process_media(uploaded_file)
+else:
+    st.warning("Please upload a file.")
+
+
+
+
+